Accumulating evidence supports the contention that the presence of disseminated Mycobacterium Avium Complex (MAC) disease contributes to morbidity and adversely impacts survival of patients with AIDS. Persons with advanced stages of HIV infection are at risk for developing MAC. This study is designed to evaluate two single-agent prophylaxis regimens and a two-drug combination regimen for prevention of MAC bacteremia and disseminated MAC disease in HIV-infected patients with CD4 counts less than 100 cells/mm3. The study opened to accrual in April 1993 and closed to accrual in January 1994. In February 1994, the study was modified to reduce the dose of rifabutin to 300 mg/day for both the monotherapy and combination therapy arms. This was based on reports of uveitis in patients participating in clinical trials of higher doses of rifabutin, and on the results of another study demonstrating a drug interaction between clarithromycin and rifabutin resulting in higher rifabutin plasma levels when the two drugs were combined. Preliminary conclusions are as follows: 1) clarithromycin alone was more effective than rifabutin alone in reducing the incidence and delaying the time to development of MAC bacteremia or MAC disease; 2) the combination of clarithromycin and rifabutin was more effective than rifabutin alone but not more effective than clarithromycin alone; 3) the time to permanent treatment discontinuation due to toxicity was significantly shorter for the combination therapy arm; 4) no survival differences were demonstrated among the three treatment arms; 5) the use of rifabutin alone did not appear to be associated with selection of rifabutin- or clarithromycin-resistant MAC isolates, whereas the use of clarithromycin alone did appear to be associated with selection of clarithromycin resistance for 27% of isolates; 6) the combination of clarithromycin with rifabutin when used for prophylaxis did not appear to significantly reduce the rate of development of clarithromycin resistance. The final data analysis for this study has not been completed.